Enabling Adaptive HMI Through Low-Interference Multimodal Sensor-Based Discrimination of Distress and Eustress
Yanzeng Zhao, Keyong Zhu, Wei Guo, Hui Gang Xu, Lijing Wang · IEEE Sensors Journal · 2025
Differentiating performance-impairing distress from beneficial eustress within a single task is a critical challenge for adaptive human-machine interaction (HMI), requiring new experimental paradigms and validated, low-interference monitoring solutions. To address this, the present study introduces a novel methodology centered on an ecologically valid experimental paradigm combined with a low-interference, multimodal sensor suite (Electrocardiogram (ECG), pupil, and voice). This methodology was validated in a flight simulation with 30 male flight trainees, where a machine learning model fusing the collected signals achieved 90% accuracy in discriminating between distress and eustress. Furthermore, the analysis explores key physiological features from each modality that indicate these distinct states. This work provides a preliminary methodology, from data acquisition to physiological insight, intended to not only support the development of next-generation adaptive HMI systems for enhanced safety and performance but also to deepen the scientific understanding of human state and behavior under stress through advanced sensing.